A comparative study of approximate entropy and sample entropy in the diagnosis of variation of runoff series：a case study in Yarkant River in Xinjiang, China
Identifying and diagnosing the aberrance point of runoff series is a key link in the reconstruction and consistency of runoff series, which is of great value for understanding hydrologic mechanism and coping with runoff changes. The measured daily runoff data of the Kaqun hydrologic station at the Yarkant River exit from 1957 to 2016 were selected. The traditional Mann-Kendall mutation test was used to identify the mutation points and the sample entropy and approximate entropy methods were used to diagnose the aberrance points of the sequence and compare the methods. The results showed that the traditional mutation test method Mann-Kendall method found multiple false mutation points, which cannot accurately determine the year of the mutation. The approximate entropy method was used to diagnose the aberrance point of the runoff sequence, and the year of 1959, 1987 and 2004 were found. For the mutation point, the sample entropy method was used to diagnose the aberrance point of the runoff sequence, and two aberrance points were found in 1987 and 2004, respectively. Through the comparative analysis of the two methods, we found that the sample entropy can better reflect the fluctuation of the runoff series than the approximate entropy in the same dimension, and the sample entropy has better sensitivity. The upstream of Kaqun Station is the alpine distribution area, the human disturbance is small, and the natural factors become the main factors affecting the runoff change in the study area.